Concept of CBIR
Content Based Image Retrieval(CBIR) is the technology to search images from image database without other information like title and author.
It is basically divided to 2 parts : Indexing and Searching. Indexing performs based on color, texture, and shape. For fast and accurate searching, developing indexing algorithm has been important issue in computer vision field.
Open Source CBIR Engines
- GPL, developed until 2006
- written in C# .NET 2.0 using Visual Studio .NET 2005
- Open license, developed until 2009
- Written in C++, Supports python
- full server system written in c, developed until 2004
- divided to kernel part and plugin part
- very complex, many dependency in other gnu projects
- developed until 2012
- server system written in python/C++/Java, used SOAP
- Recently developed, GPL
- written in Java, use OpenCV, LSH
- business model : source code customization consulting
- image recognition platform for your mobile apps
- Recently developed. LGPL
- Written in C++, consist of 10 cpp files. use OpenCV 2.4
- Include HTTP server
- Used ORB instead of SIFT or SURF (patent protected)
- business model : Image database server hosting, source code customization
- Pastec can be a good reference to make a content based image retreval system because it is simple and has major features. Also it is free from license/patent.
- However, it is wired that they are promoting that it is for mobile apps. Intensive test is needed.